ZeroGPU fast mode: skip delight model, fewer camera views, lower view size, 8 denoise steps

Reduce texgen resolution to 1024 and enable CPU offload to lower GPU memory pressure

Add explicit runtime diagnostics and exception logging for GPU execution

Prefetch Hunyuan weights at startup to keep GPU task within duration budget

Use Hunyuan3D-2.1 custom_rasterizer wheel; restore ZeroGPU-supported torch 2.10

Load torch before custom_rasterizer import to resolve libc10 dependency

Try torch 2.5.1 + tencent custom rasterizer wheel; simplify rasterizer bootstrap

Build custom_rasterizer from source at runtime with discovered CUDA paths

Try Step1X custom_rasterizer wheel for torch ABI compatibility

Pin torch/torchvision/diffusers/transformers to CUDA12-compatible stack

Fix runtime CUDA shared library path and lazy imports

Fix custom_rasterizer wheel URL to source space repo

Fix build: install custom_rasterizer wheel from HF URL

Initial Space: image+GLB attire transfer with Hunyuan3D-2 texgen (ZeroGPU 60s)

initial commit